This proposal is in response to PA 18-484. In the BRAIN 2025 Report, PET (positron emission tomography) is identified as ?the best means to translate studies of neurotransmitters, receptors, and neuromodulators to humans.? However dynamic assessment of receptor occupancy and metabolism is hindered by the spatial resolution and sensitivity of even the most modern of clinically available PET scanners. To address this challenge, we propose a next generation PET detector with a highly innovative design: a detector module with a layered scintillator structure and a side readout configuration. The crystal slabs in the module are stacked along the depth direction and are optically separated by reflective films. The scintillation light created in each layer is measured by photodetectors located on the four sides of the crystal. Compared with traditional PET detectors, which contain pixelated crystal arrays, the new design has the following advantages: (1) The layered structure provides depth of interaction (DOI) information such that a smaller diameter detector ring can be used without increasing parallax error, increasing sensitivity while lowering costs. (2) The four-sided readout method improves the energy resolution of the system with increased scintillation light collection efficiency by reducing light loss due to total internal reflection. (3) Sub millimeter spatial resolution is achievable without using very small pitch crystal arrays, since the interaction location in each crystal layer is determined via machine learning- based decoding of the light distribution collected on the four crystal sides. Therefore the production cost of the crystals is reduced. (4) Since the interaction location and energy resolution for each layer are determined independently, the system sensitivity can be increased by stacking more layers in the module without affecting the spatial and energy resolution of the system. (5) For side readout setup, a larger ratio of cross-sectional area to length requires fewer photodetectors to cover all four sides of the module; this reduces photodetector cost. The first four points above have been demonstrated in preliminary studies using a small, prototype module. We propose to build a large scale detector module with this new design to verify the fifth advantage, and to study the effect of detector size on the first four. The outcome of this proposal will be two DOI enabled detector modules with excellent spatial resolution (~1 mm) and energy resolution (~10%), as well as good timing resolution (~400 ps), and DOI resolution (~ 3 mm) and high system sensitivity. A full characterization study for the two modules and imaging studies for both the Derenzo and Hoffman brain phantoms will address the current limitations of human brain PET scanners, and will serve as the foundation for a new dynamic PET scanner for neuroimaging.